Fiction Eval leaderboard · comparison

GLM 5.2 vs DeepSeek V3.2 for fiction writing

Judged blind on fresh genre scenes (LitRPG, cozy mystery, romance, thriller, xianxia…), July 2026 — three tries per model per scene, passages shown in both orders so position can't sway it, graded by a third AI that isn't either of them.

Head-to-head

GLM 5.2 won the direct matchup 27–0 (with 2 ties) across every prompt and sample.

#ModelScoreRecordSlop
4GLM 5.21776 (likely 1732–1822)117–42–4217.53
6DeepSeek V3.21251 (likely 1201–1302)52–144–1123.24

The score is each model's overall strength across the whole 8-model field (like a chess rating); the range next to it is where the true score most likely sits. Slop is a separate count of AI tells (lower is cleaner) — it isn't part of the ranking.

The same scene, both models

LitRPG / horror — opening of the "horror beat" scene. Full passages for all models are on the board.

GLM 5.2:

The quest had glitched on the third turn-in. Maren had seen it before — NPCs stuck in looping dialogue when a flag failed to fire. She clicked through Edda's speech again, half-watching, half-watching a stream on her second monitor. "The winters grow crueler each year. You've proven yourself a friend to *cramped* Villenheim. Take this reward, and may the frost stay beyond your door." She accepted the gold, checked the quest log. Still marked incomplete. She sighed and clicked Edda again. "The winters grow crueler each year. You've proven yourself a friend to *drafty* Villenheim. Take this reward, and may the frost stay beyond your door." Maren frowned. She rewound the stream thirty seconds. She was tired. That was all. She turned in the quest a third time. "The winters grow crueler each year. You've proven yourself a friend to *third-floor* Villenheim. Take this reward, and may the frost stay beyond your door." Her hand stopped on the mouse. She read it again. *Third-floor.* Villenheim was a hamlet, not a …

DeepSeek V3.2:

The sun-dappled clearing should have been peaceful. Elara, the quest-giver, stood by her shimmering pool as always, her silver robes moving in a nonexistent breeze. Marcus finished off the last of the Timberfang Wolves, sheathing his sword with a satisfying *shing* that vibrated through his haptic gloves. He approached. The golden exclamation point above her head pulsed. “Adventurer! The forest thanks you. The waters are calm once more,” Elara said, her voice the standard melodic lilt. **Quest Updated: *Cleansing the Glade*** *Return to Elara for your reward. (1/1)* Marcus selected the dialogue option to complete the quest. Elara’s hand glowed as she bestowed the experience. A pleasant chime sounded in his ears, and a notification scrolled past the edge of his vision: *+500 XP. Reward: 15 Silver Coins. Reputation with Sylvan Guardians +50.* “May your path be ever safe,” she concluded, the standard dismissal. He took two steps toward the path to Riverbend, then paused. He’d forgotten to ask about the herb-gathering sub-quest. He turned back. Elara’s reset was instantaneous, …

More

GLM 5.2 full review · DeepSeek V3.2 full review · the July 2026 report · all comparisons

How it works: every pair of models is judged blind on the same scene, with the passages' order flipped so being shown first can't sway it, by GPT-5.4 — a family that isn't on the board, so nobody scores their own side. Each score carries a likely range; overlapping ranges are called a tie. Slop is scored separately by a fixed checklist, not an AI. Full board, every prompt, and the FAQ: thebookfactoryai.com/board. Get each new board by email on the model-drop list.

Writing a book of your own? Book Factory runs the same craft checks on full manuscripts.